基于SVDD的冷水机组传感器故障检测及效率分析

摘要

In the refrigeration and air conditioning system, sensors are independent component for physical data measuring and operating state monitoring. Sensor faults, especially sensor biases output will lead to incorrect measurement, inappropriate controlling strategy and further energy consumption rise. Based on the pattern recognition theory, the fault detection task could be considered as a one-class classification problem. Therefore, a powerful pattern recognition-based one-class classification algorithm, Support Vector Data Description (SVDD) was used to detect the sensor biases occurring in a chiller system. The practical fault-free data were used as training dataset to develop the SVDD model so as to detect the sensor faults. The method and its fault detection efficiency were validated by test dataset with different artificially introduced levels of sensor biases. The SVDD-based fault detection method worked well with chiller practical operating measurements, but the fault detection efficiencies of different sensors with different level faults were inconsistent.%传感器是制冷空调系统的重要组成部分，起着测量数据和监控状态的作用。传感器故障，尤其是输出偏差会引起测量值不准，影响控制策略，导致系统能耗增加。依据模式识别理论，故障检测可处理为一种单分类问题。据此采用一种单分类模式识别工具——支持向量数据描述（SVDD），针对冷水机组进行了偏差故障条件下的传感器故障检测工作。收集冷水机组实测正常运行数据，基于训练集建立SVDD模型，进行冷水机组传感器故障检测；在测试集中引入不同幅值水平的偏差故障，分析检测效率。结果表明：基于SVDD的冷水机组传感器故障检测效果明显，但对于不同传感器的不同幅值偏差故障，故障识别程度并不一致。